Would You Trust a Chatbot?

Abhinav Raj, Writer
@uxconnections

As advances in ML, AI and natural language processing (NLP) turn chatbots more humanlike, should companies tell customers they are communicating with bots?

From little gimmicks put in place for embellishment to indispensable tools on commercial websites, chatbots have come a long way. 

With 23% of customer service organizations employing AI-enabled chatbots, the little assistants have occupied the bottom-right corner of every fifth website to become its de-facto concierge. 

But are chatbots any good at solving problems? 

It appears so. 

According to Forbes, out of the 60% of millennials who have used chatbots, 70% reported positive experiences at the end. The bots offered the customers instant gratification through conversational engagement—while taking a significant load off the shoulders of customer service executives by reducing call, chat and email enquiries. 

With the onset of natural language processing (NLP) technology, chatbots have become more human-like than ever before, whilst simultaneously becoming better at solving problems. With the advent of deep learning, businesses can deploy NLP-based chatbots that are better at assessment, analysis and clear and coherent communication. 

As chatbots become more indistinguishable from their human counterparts, should companies tell customers that they’re talking to machines? 

Yes and no, according to researchers at the University of Göttingen. 

In a recent survey conducted by the university, 400 participants were asked to contact their energy providers with a simple objective—to update the address on their electricity contract. Out of the 400 participants, half were informed that they would be conversing with a chatbot while the other half remained blissfully unaware.

The research gauged the impact of this disclosure based on the chatbot’s ability to find a resolution, and how important the customer’s perception of the said resolution turned out to be. A plethora of scientific methods such as covariance and mediation analysis were employed in the study. 

It was found that, predominantly, the reaction of the customer was negative upon the revelation that the conversational partner is a chatbot, and this particular scenario weakened customer trust. Paradoxically, however, the disclosure has a positive impact on customer reactions in cases where the chatbot is unable to offer a meaningful resolution.

“If their issue isn’t resolved, disclosing that they were talking with a chatbot, makes it easier for the consumer to understand the root cause of the error,” notes the first author of the study, Nika Mozafari. 

“A chatbot is more likely to be forgiven for making a mistake than a human.”

The study appears in the Journal of Service Management. 

The key takeaway is that while chatbots have been improving, the general notion of the public remains apprehensive towards the technology. However, provided the advancements in NLP and ML algorithms that run modern chatbots make them virtually indistinguishable from humans, it may not be a good idea to name your chatbot something like… Sir Chatsalot.